Dingwen Yuan

Instrumenting Wireless Sensor Networks A survey on the metrics that matter

Wireless Sensor Networks (WSN) are part of the technical fundament enabling the Internet of Things (IoT), where sensing and actuator nodes instantaneously interact with the environment at large. As such they become part of everyday life and drive applications as diverse as medical monitoring, smart homes, smart environment, and smart factories, to name but a few. To acquire data, individual sensors interact with the physical environment by sensing physical phenomena in proximity. The wireless network connectivity is leveraged to collect the raw data or pre-processed events, and to disseminate code, queries or commands. Actuating capabilities facilitate instant interactions with the environment or application processes. Experience on how to operate large scale heterogeneous WSNs in (critical) real-world applications is still scarce, and operational considerations are often an afterthought to WSN deployment. A principled look into the metrics, i.e., a standard or best practice of measurement of the vital parameters in WSNs is still missing. In this article, we contribute a survey on the most important metrics to characterize the performance of WSNs. We define an abstract system model for WSNs, take a look on what the WSN community considers metrics that matter, and categorize the metrics into scopes of relevance. We discuss the properties of the metrics as well as practical aspects on how to obtain and process them. Our survey can serve as a manual for implementors and operators of WSNs in the IoT.

Let's talk together: Understanding concurrent transmission in wireless sensor networks

Wireless sensor networks (WSNs) are increasingly being applied to scenarios that simultaneously demand for high packet reliability and short delay. A promising technique to achieve this goal is concurrent transmission, i.e. multiple nodes transmit identical or different packets in parallel. Practical implementations of concurrent transmission exist, yet its performance is not well understood due to the lack of expressive models that accurately predict the success of packet reception. We experimentally investigate the two phenomena that can occur during concurrent transmission depending on the transmission timing and signal strength, i.e. constructive interference and capture effect. Equipped with the thorough understanding of these two phenomena, we propose an accurate prediction model for the reception of concurrent transmission. The extensive measurements carried out with varying number of transmitters, packet length and signal strength verify the excellent quality of our model, which provides a valuable tool for protocol design and simulation of concurrent transmission in WSNs.

Ripple: High-throughput, reliable and energy-efficient network flooding in wireless sensor networks

The recently proposed Glossy protocol demonstrated the high potential of constructive interference (CI) in improving communication performance in wireless sensor networks. This paper presents a network flooding protocol, Ripple, which also exploits CI while improving Glossy in terms of throughput and energy efficiency by a factor of three each. To this end, we propose to pipeline transmissions on multiple channels. Ripple uses a novel packet-based channel assignment to eliminate the time overhead occurring in traditional node-based channel assignment procedures. Moreover, if we apply the Reed-Solomon (RS) erasure code to Ripple, it pushes the reliability close to 100%, surpassing Glossy, being nonetheless computationally practical for TelosB motes. Still, the throughput after error coding doubles, or even triples that of the state-of-the-art reliable data dissemination protocol Splash. By tuning the transmission interval, Ripple balances between high throughput and high reliability, thus suiting an array of network broadcast applications with various QoS requirements.

Tree-based multi-channel convergecast in Wireless Sensor Networks

The use of Wireless Sensor Networks (WSN) in application domains such as industrial automation poses strict requirements in terms of communication delay, reliability, etc. Time Division Multiple Access (TDMA) is the preferred access scheme for the above applications, and leading industry standards such as WirelessHART employ it by centrally generating convergecast schedules. Despite the fact that techniques such as multi-channel communication and spatial re-use are currently not harnessed to their full potential, existing scheduling heuristics are often complex while performing far from the optimal solution. We investigate the tree convergecast scheduling with multiple channels (TCMC) problem. In particular, we derive an integer programming-based optimal solution to the min length and buffer size scheduling as well as the min length and channel number scheduling. We further describe TCMC scheduling as a decision problem, which allows us to create a general scheduling framework that is flexible and requires minimal code modification for implementing different TCMC scheduling strategies. Within our framework, we propose and implement four heuristics. Our novel busy-sender-first heuristic is significantly better than the state-of-the-art heuristic in both schedule length (within 0.22% of the optimum) and memory consumption, as well as being conceptually much simpler. Finally, based on the evaluation results of the busy-sender-first heuristic, we derive guidelines on the choice of number of channels and configuration of tree topology, respectively.

HOPSCOTCH: An adaptive and distributed channel hopping technique for interference avoidance in Wireless Sensor Networks

Interference is one of the key factors impacting the performance and robustness of Wireless Sensor Networks (WSN) operation. For serious WSN applications, it is crucial that interference mitigation acts in a fast and reliable fashion. We describe an adaptive and distributed channel hopping scheme, HOPSCOTCH1. Our scheme is novel in the sense that (1) it is built on a lightweight yet accurate metric to describe the interference, and (2) it is fully distributed in nature and combines a proactive, consent-based as well as a reactive, rendezvous-based hopping technique, which allow for robust operation even in adverse conditions. We show by extensive experimentation that our channel metric models real-world conditions accurately and that HOPSCOTCH provides a very fast response time to adapt the network to interference.

Patrolling wireless sensor networks: randomized intrusion detection

Wireless sensor networks typically consist of highly resource-constrained motes. Hence, it is desirable to reduce the tasks of each mote to a minimum. We claim that even critical security functions such as intrusion detection can be performed by means of randomizing the detection frequency with the goal of making it more lightweight. To this end, we present Patrol, a system which distributes the load caused by various tasks across the network. Patrol makes use of tokens that are exchanged between nodes and activate a certain functionality, such as intrusion detection, temporarily. As a proof-of-concept, we design and implement within Patrol a lightweight intrusion detection algorithm based on the energy consumption of the nodes. We show that by analyzing the energy consumption, flooding attacks can be detected reliably. To illustrate these facts, we use a real-world testbed consisting of the widely-employed TelosB motes.

A secure monitoring and control system for Wireless Sensor Networks

The maintenance of Wireless Sensor Networks (WSNs) can carry high or prohibitive costs, particularly, if the WSN is deployed in unattended areas. Secure monitoring and control of the WSN is vital, however, practical systems are rare and limited with respect to their capabilities. We present a monitoring and control system for WSNs that is secure and additionally equipped with intrusion detection functionality. It allows to reliably assess the actual status of the network, to configure the sensor nodes, and to further use this data to highlight suspicious events.